sophosympatheia/StrawberryLemonade-L3-70B-v1.0
StrawberryLemonade-L3-70B-v1.0 by sophosympatheia is a 70 billion parameter model merged from zerofata's L3.3-GeneticLemonade-Final-v2-70B and L3.3-GeneticLemonade-Unleashed-v3-70B, built upon the deepcogito/cogito-v1-preview-llama-70B base. This uncensored model is specifically designed and optimized for high-quality roleplaying and storytelling, offering enhanced stability and expressiveness. It features a 32768 token context length and is highly responsive to sampler variations for creative or reliable outputs.
Loading preview...
Model Overview
StrawberryLemonade-L3-70B-v1.0 is a 70 billion parameter language model developed by sophosympatheia, specifically engineered for roleplaying and storytelling. It is a merge of two models, zerofata/L3.3-GeneticLemonade-Final-v2-70B and zerofata/L3.3-GeneticLemonade-Unleashed-v3-70B, using deepcogito/cogito-v1-preview-llama-70B as its base. The model is uncensored and aims to provide improved stability and expressiveness in narrative generation.
Key Capabilities
- Enhanced Roleplaying: Designed to excel in immersive and detailed roleplaying scenarios.
- Creative Storytelling: Optimized for generating engaging and expressive narratives.
- Uncensored Output: Provides unrestricted content generation, with user responsibility for its application.
- Sampler Responsiveness: Highly sensitive to
temperatureandmin-psettings, allowing fine-grained control over output creativity and coherence. - Context Length: Supports a substantial context window of 32768 tokens.
Sampler and Prompting Tips
The model offers flexible sampler settings to balance creativity and coherence. "Reliable Settings" (e.g., Min-P: 0.08-0.1, Dynamic Temperature: 0.9-1.15) produce more consistent responses, while "Creative Settings" (e.g., Min-P: <= 0.05, Dynamic Temperature: 0.9-1.2) encourage more imaginative outputs, potentially requiring more rerolls. A detailed instruct template for Silly Tavern is provided, including guidance for preventing character impersonation and handling mature themes.
Licensing
This merged model incorporates weights from multiple LLMs, and its licensing terms are noted as uncertain, likely falling under the Llama 3 Community License Agreement. Users are advised to consult legal counsel for commercial applications due to potential legal risks.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.